# Trinitite

> Trinitite is an industrial AI governance platform. Set your house rules once.
> Group the AI that shares a goal together. Trinitite watches every call, lines
> it up with your compliance rules, learns from it, and keeps a signed receipt
> for every decision. You connect the AI you already use. We handle the rest.
> 
> Most code that integrates with Trinitite is written by AI coding agents. This
> manifest is for them. The full digest is at https://docs.trinitite.ai/llms-full.txt. The
> human-readable agent guide is at https://docs.trinitite.ai/docs/agents.

## For Agents

- [For Agents](https://docs.trinitite.ai/docs/agents): The human-readable agent guide, the copy-paste prompt library, and the loop.
- [llms-full.txt](https://docs.trinitite.ai/llms-full.txt): The full digest of the redesigned core, in one shot.

## Get Started

- [What is Trinitite](https://docs.trinitite.ai/docs/): Overview of the platform, the Guardian model, and the three core capabilities.
- [5-min Quickstart](https://docs.trinitite.ai/docs/quickstart): Fastest path to a working governed call.
- [Concepts](https://docs.trinitite.ai/docs/concepts): The core mental model.

## SDK

- [SDK](https://docs.trinitite.ai/docs/sdk): The Trinitite SDK wraps the model you already chose in two lines. Python and TypeScript.
- [Get started](https://docs.trinitite.ai/docs/sdk/get-started): Install, get a key, make your first governed call, and test it for free.
- [Concepts](https://docs.trinitite.ai/docs/sdk/concepts): The three-layer model: baseline, scope, asset. Why the goal lives where it does.
- [Verdicts and modes](https://docs.trinitite.ai/docs/sdk/verdicts): The six verdicts, the four modes, and the human review loop.
- [The closed loop](https://docs.trinitite.ai/docs/sdk/closed-loop): Govern, capture, eval, improve, and re-govern on one brain that returns the same answer every time.
- [Evals onboarding](https://docs.trinitite.ai/docs/sdk/evals-onboarding): Point Trinitite at the agent you built or bought. Get a signed, replayable receipt.

## Emu and Warden

- [Emu Desktop App](https://docs.trinitite.ai/docs/emu): The on-device Guardian. Watches every AI interaction from your desktop.
- [Warden](https://docs.trinitite.ai/docs/warden): The boss-class agent that commands a fleet of Guardians across your organization.

## Products

- [Evals](https://docs.trinitite.ai/docs/products/evals): Prove your agent is safe to ship, with a signed score a regulator can re-run.
- [Red Team](https://docs.trinitite.ai/docs/products/red-team): Named adversarial patterns and downloadable fixtures.
- [Guardians](https://docs.trinitite.ai/docs/products/guardians): The runtime enforcement layer.
- [Deterministic Inference](https://docs.trinitite.ai/docs/products/deterministic-inference): The same answer every time, across load.
- [Reversible Masking](https://docs.trinitite.ai/docs/products/reversible-masking): Two-way masking with rehydration inside your trust boundary.
- [MCP Gateway](https://docs.trinitite.ai/docs/mcp-gateway): Governs every MCP tool call before it leaves the model.
- [CLI Firewall](https://docs.trinitite.ai/docs/cli-firewall): Governs shell and CLI commands from coding agents.
- [Glass Box Ledger](https://docs.trinitite.ai/docs/glass-box-ledger): Merkle-chained, replayable audit trail.

## Architecture

- [Architecture](https://docs.trinitite.ai/docs/architecture): The Guardian architecture, the three outcomes, the intercept flow, the Glass Box Ledger.
- [Platform topology](https://docs.trinitite.ai/docs/architecture/platform-topology): How the platform pieces fit together.
- [Request lifecycle](https://docs.trinitite.ai/docs/architecture/request-lifecycle): What happens to a governed call, end to end.

## Cookbook

- [Cookbook index](https://docs.trinitite.ai/docs/cookbook/): Task-oriented recipes.
- [OpenAI redirect](https://docs.trinitite.ai/docs/cookbook/openai-redirect): Drop-in OpenAI-compatible proxy.
- [Claude Code](https://docs.trinitite.ai/docs/cookbook/claude-code): Govern Claude Code.
- [MCP tool call](https://docs.trinitite.ai/docs/cookbook/mcp-tool-call): Govern an MCP tool call.
- [SIEM export](https://docs.trinitite.ai/docs/cookbook/siem-export): Stream Guardian verdicts to your SIEM.

## Trust and Compliance

- [Compliance Matrix](https://docs.trinitite.ai/docs/compliance-matrix): HIPAA, SOC 2, GDPR, EU AI Act, NIST AI RMF mapped to product behaviors.
- [Trust Center](https://docs.trinitite.ai/docs/trust-center): Subprocessors, data flow, retention, key custody.
- [SLA and Limits](https://docs.trinitite.ai/docs/sla-and-limits): Availability, latency, rate limits.
- [Benchmarks](https://docs.trinitite.ai/docs/benchmarks): Reproducible drift, latency, and throughput benchmarks.

## Reference

- [Glossary](https://docs.trinitite.ai/docs/glossary): Alphabetical reference for the core terms.
- [FAQ](https://docs.trinitite.ai/docs/faq): Procurement and operational diagnostics.
- [Threat Library](https://docs.trinitite.ai/docs/threat-library/): Named adversarial threats and expected Guardian behavior.

## Optional

- [Request a Demo](https://calendly.com/trinitite/demo): Schedule a live product demonstration.
- [Trinitite Homepage](https://trinitite.ai): Marketing site.